Closed adelavega closed 1 year ago
Base: 75.11% // Head: 75.27% // Increases project coverage by +0.15%
:tada:
Coverage data is based on head (
d92c637
) compared to base (ecdf208
). Patch coverage: 97.61% of modified lines in pull request are covered.
:umbrella: View full report at Codecov.
:loudspeaker: Do you have feedback about the report comment? Let us know in this issue.
TFHubAudioExtractor
This extractor is very similar toTFHubImageExtractor
but with 1) slightly different preprocessing 2) infers timing output of model (i.e. onsets, durations) from the output (assumes fixed sampling rate)saved_model
loading toTFHubExtractor
. Models likeSPICE
are not meant to instantiated asKerasLayer
. Instead, they must be loaded withhub.load
and have predefinedsignature
where the model is stores, and named vectors as outputs.For reference this is the
SPICE
model: https://tfhub.dev/google/spice/